You are on page 1of 22

MATH 6342 - Advanced Enumeration

Final Project - Automorphism groups and permutation groups


Tuesday, December 14, 2021

Name: Juan F. Rodriguez-Quinche M.U.N Number: 202093619

Definitions of a group
The definition of a group has had an historical path going from that has changed
a lot through the years. We will show some equivalent definitions of this:

• The first definition of a group, as we said before, started as “all symmetries


of an object”, and as object we mean a pair (X, S) where X is a set and S is
a structure on X, for example a graph, a set of partition or something even
more complex. For example, lets consider the following graph:

Figure 1: Example Graph

If we want to list the possible symmetries of this object (X, S), (this is the
set of vertices and edges, together with in an structure of a graph) note the
following:

– The ‘do nothing’ symmetry is considered always.


– Permutation of the 3 leftmost edges, give us 6 different symmetries. We
can call this Sym3 .
– A reflection through an horizontal axis give us 2 symmetries. We can
call this Sym2 .
– The action of permuting edges and reflecting through an horizontal axis
is commutative. Hence we can consider the symmetries of this graph

1
to be Sym3 × Sym2 , this means, each symmetry corresponds to a com-
bination of both, a permutation on the leftmost edges and a reflection
on an horizontal axis.

The point of this definition is that, given a permutation, there is a natural


way to apply it to the object. In our example, we will name a element of the
symmetries as g, and we can apply g to the graph to obtain gS, that we can
think it as the same elements of X but with the structure of gS, and those
two graphs are equivalent.

• For the second definition, we need a couple of extra definitions:

1. A permutation g on X is an automorphism if gS = S.
For example, if we permute the rightmost vertices of Figure 1 but not
the rest of them, this will not preserve the initial graph structure. Hence
that would not be an automorphism.
2. The automorphism group of (X, S) is the set Aut(X, S) of all automor-
phisms of (X, S).
3. A subset G of the symmetries of X is an automorphism group if G =
Aut(X, S) for some structure S on X. This was the first definition of a
group for us.

Consider the following properties of an automorphism group:

(P1): it contains the identity permutation,


(P2): it contains inverses of each element,
(P3): it contains the composition of each pair of elements.

Note that this are true because:

1. The first condition is clear.


2. If we got a permutation g on X such that gS = S and we apply g −1 to
both sides, we get g −1 gS = g −1 S so g −1 S = S, hence g −1 is an element
of the automorphism group.
3. If we got two permutations g, h, by definition gS = hS = S, hence
(hg)S = h(gS) = (h)S = S.

And our second definition is given by this three properties. We say that a
set G is a permutation group on X if G satisfies (P1), (P2) and (P3).
Is clear, due to the definition, that every automorphism group is a permuta-
tion group. The other direction is given by the following:

2
Theorem 1 (Exercise 14.8.1). Every permutation group is the automorphism
group of some object.

Proof. Let’s consider a permutation group G on A = a1 , . . . , an we have to


exhibit a set X and an structure S such that Aut(X, S) = G. This set will
be exactly X = A and S = G, i.e. Aut(A, G) = G.
Let’s consider the permutation g ∈ G to n-tuple (a1 , . . . , an ), this is exactly
(ga1 , . . . , gan ) and this new n-tuple is equivalent to g as G is a permutation
group. Let’s consider the action of h ∈ G to this new n-tuple, note that
this is (h(ga1 ), . . . , h(gan ) = ((hg)a1 , . . . , (hg)an , i.e. G is permutating the
elements of g itself as a group of permutations while preserving its structure,
this implies that Aut(A, G) = G.

This implies that the two definitions are equivalent.


• Now we will describe the axiomatic definition, for this, we will denote the
identity permutation as e, and as the identity we got that, for any other
permutation g,
eg = ge = g,
and its inverse will be denoted as g −1 , and,

gg −1 = g −1 g = e,

finally, the composition between g and h we will write it as g ◦ h. Note that


a permutation group follows this properties:
(A1): Associativity: g ◦ (h ◦ k) = (g ◦ h) ◦ k, for all g, h, k ∈ G,
(A2): Identity: there exists e ∈ G such that e ◦ g = g ◦ e = g, for all g ∈ G,
(A3): Inverses: for any g ∈ G, there exists g −1 ∈ G such that g ◦ g −1 =
g −1 ◦ g = e.
In the case of a permutation group over X note that the associativity is a
general property of functions:

x(g ◦ (h ◦ k)) = (xg)(h ◦ k) = ((xg)h)k = (x(g ◦ h))k = x((g ◦ h) ◦ k).

We define an abstract group G with binary operation ◦ as a set that fulfill


properties (A1), (A2) and (A3).
Note that the inverse and identity permutations have the properties men-
tioned, and are contained in a group G, and moreover, every permutation
group is an abstract group, for the other direction we got the following the-
orem:

3
Theorem 2 (Cayley’s Theorem). Every abstract group is isomorphic to a
permutation group.

Proof. Let G be an abstract group, we want to find a permutation group G′


over a set X, whose elements are in a correspondence to the elements of G
and such that the composition of g and h corresponds to g ◦ h; this is the
definition of being isomorphic.

Let’s take X = G and, let ρg be defined as:

ρg x = g ◦ x,

this is the right translation by g ∈ G, moreover, let’s take G′ = {ρg : g ∈


G}.Note that if h ̸= g then ρh ̸= ρg , due to the fact that, if we let them act
on e ∈ X we got:

ρh e = h ◦ e = h ρg e = g ◦ e = g,

this implies a one-to-one correspondence between the elements of G and the


elements of G′ .

It isn’t clear that ρg is a permutation, but:

ρg ρh x = (g ◦ h) ◦ x = g ◦ (h ◦ x) = ρg◦h x,

this tells us that the group operation of G corresponds to the composition


of the permutations, moreover, ρe is the identity permutation and ρ−1 g is the
inverse of ρg . It follows that properties (P1), (P2) and (P3) are fulfilled, and
indeed G′ is a group of permutations over X.

It also follows that every abstract group is also an automorphism group.

Now is clear that our three definitions are equivalent, but moreover, Frucht showed
that every abstract group is the automorphism group of a graph, this will be
approached further.
From now and on, we abbreviate ‘abstract group’ to ‘group’ and g ◦ h to gh. A
group will be defined by the axioms (A1)-(A3), and note that Cayley’s theorem
imply that it makes sense to study groups as groups of permutations, but note
that the definition of ‘permutation group’ changes then, now we will take it as a
set of permutations which is equipped with the composition of these, and it makes
a group.

4
Now we will define the action of a group on a set X a map θ : G → Sym(X)
that satisfies the following:

1. θ(gh) = (θg)(θh),

2. θ(1) = (1),

3. θg −1 = (θg)−1 .

Here we use the same notation for a group operation and permutations, more-
over, θ is an homomorphism from G into Sym(X).
Is also interesting to note that a group can have many different actions, so we need
to find a way to tell when two actions are the same, or equivalent. For this, let’s
consider two actions θ and ϕ from G into sets X and Y respectively. We will say
that these actions are equivalent if there exists a bijection f : X → Y such that,
for any x ∈ X and g ∈ G:

(ϕg)(f (x)) = f (x(θg)x),

this is if f give us a correspondence between X and Y , any element of G induces


the same permutation in their corresponding sets.
For example consider the following graph:

a b

We define G = S3 to be the automorphism group of a triangle, we can define


two actions of G on the triangle, one on the vertices and another one on the edges.
And let f be the map defined as:

{b, c}
 if x = a
f (x) = {c, a} if x = b

{a, b} if x = c

This maps show that both actions are equivalent, for example if a permutation
g ∈ G send a to b then it also carries {b, c} to {c, a}.

5
Examples of groups
Lets consider the following additional definitions:
• A subset H ⊂ G is a subgroup if H is closed under the operation of G.

• The number of elements a group has is the order of a group.

• A subset S ⊂ G is a generating set of G if any element g ∈ G can be


expressed as the product of the elements of S and its inverses
Note that, due to the definition of a generating set of G, it can not be contained in
any proper subgroup of G, but more generally, if S is a subset of G, the subgroup
H, generated by S consists on all the products of elements of S and its inverses.
And it can be also characterized as the smallest subgroup that contains S.

A finite group (group with finite order) that can be generated by only one
element is called cyclic group and denoted Cn . The group Cn can be regarded in
many different ways, for example:
• The additive group of congruence classes modulo n.

• The multiplicative group of all nth roots of the unit in C, i.e. {e2kπi/n : k =
0, . . . , n − 1}.

• The automorphism group of the following digraph, for n > 2:

1
n 2
3
···

10 4

9 5
8 6
7

Figure 2: Cyclic digraph

Another example of groups, that is closely related to cyclic groups, are dihedral
groups, D2n of order 2n. We can think on the dihedral group for n > 2 to be
either:

6
• The automorphism group of the undirected graph of figure 2.
• The group of symmetries of a regular n-gon.
Is easy to note that Cn is contained as a subgroup of D2n (as the rotations of
the n-gon). The remaining elements are reflections of the n-gon through n axis of
symmetry.

Also, we have already mentioned the group of symmetries of a set X, if |X| = n,


we denote this group as Sn and its order is n!. Also, this group has naturally a
subgroup only consisting on the even permutations (an even permutation is a per-
mutation that can be represented by an even number of inversions); this subgroup
is denoted as An and known as alternating group.

There are several examples of other groups, but something interesting about
we can construct new group using smaller ones. We are going to describe 3 con-
structions.
Direct product: Let’s consider two groups G and H, we define the new group
G × H, for elements being pairs (g, h) such that g ∈ G and h ∈ H. and
operation defined as
(g1 , h1 )(g2 , h2 ) = (g1 g2 , h1 h2 ).

If G and H acts on sets X and Y respectivelly, there are different actions we


can define on X ∪ Y , for future examples, we will show two,

1. The first one is known as the natural action of G × H over X ∪ Y ,


defined as: (
gz if z ∈ X
(g, h)z =
hz if z ∈ Y.
2. The second one is known as the product action, defined as:
(g, h)(x, y) = (gx, hy).

Taking all these in consideration, is easy to see that the group S3 × C2 is the
group of symmetries of the graph on Figure 1.
Wreath product The abstract definition of the wreath product between two
groups G, H, denoted as G ≀ H need more work, but we can also give a
description as a permutation group acting on X × Y , moreover, if Y =
{y1 , . . . , yn }, we can think on X × Y as n copies of X, denoted as Xi =
X × {yi }. And we will define two permutation groups:

7
u

Figure 3: Regular Octahedron

• The bottom group B is defined as G × · · · × G with the natural action


| {z }
n-times
over X1 ∪ · · · ∪ Xn , i.e.

(g1 , . . . , gn )(x, yi ) = (gi x, yi ).

• The top group T defined as the action on the second coordinate, i.e.

h(x, yi ) = (x, hyi ).

Note that the action on the second component shifts the Xi . And the wreath
product G ≀ H is the group generated by T and B i.e. consists in all the prod-
uct of elements t ∈ T and b ∈ B.

An example of a group with this construction is the group of symmetries of


the regular octahedron:

Example 3 (Exercise 14.8.2). The symmetry group of the regular octahedron


is the wreath product S2 ≀ S3 .

Proof. Let Γ be the vertices and faces of an octahedron, as in Figure 3, the


vertices of Γ can be partitioned into three pairs of opposite vertices. These
pairs are: the top–bottom pair X1 (containing the vertex labelled u; the

8
front–back pair X2 (containing v); and the left–right pair X3 (containing
w). Using reflections one can exchange the two vertices in the top–bottom,
front–back, or left–right pair, leaving the other four vertices fixed. Is very
clear that this reflection commute, so for the symmetries of Γ we know that

(S2 )3 = S2 ⊕ S2 ⊕ S2 < Sym(Γ)

Moreover, the action of a permutation on the set of vertices given by this sub-
group of Sym(Γ) is exactly the natural action, as it fixes two pairs of vertices.

Now, we want permute the each one of 8 faces of the octahedron to the
face delimited by the vertices {u, v, w}, for this we can do it by rotating the
octahedron over the middle plan or turning it upside down, for a total of 6
possible permutations, and as those permutations doesn’t commute we know
that the group that acts on the faces by permutations is exactly S3 < Sym(Γ).

Now, the action of S3 on the vertices, is exactly shifting the sets of pairs of
vertices, moreover, this is the product action. This implies that S 2 is the
bottom group and S3 is the top group, i.e. S2 ≀ S3 < Sym(Γ).

The group described above is in fact the entire group of symmetries. To


show this, let α ∈ Sym(Γ) be an arbitrary symmetry. Once we know where
α moves the vertices u, v, and w, we have determined α. Since the action
of S2 ≀ S3 is vertex transitive, there is an a ∈ S2 ≀ S3 that takes u to α(u).
Moreover a−1 α is a symmetry that fixes u. If a−1 symmetry that fixes u.
If a−1 α exchanges the front-back pair of vertices with the left-right pair of
vertices, then there is a b ∈ S2 ≀ S3 that fixes the top-bottom pair of vertices
and also exchanges the front–back pair of vertices with the left–right pair of
vertices. Taking b to be the identity if a−1 α did not make this exchange, one
sees that b−1 a−1α fixes u and takes the front-back pair of vertices to itself,
and similarly with the left-right pair. In the end we can find an element on
S2 ≀ S3 that multiplied by α fixes u, v, and w. But then this multiplication
is exactly the identity and, hence α ∈ S2 ≀ S3 .

9
Orbits and Transitivity
The main interests of groups from the algebraic perspective are the groups it-
self, but from the point of view of combinatorics, its more interesting to see the
structures X on which a group G acts on; for example what structures on X are
invariant by the action of G.

For this, consider the following relation. Let G acts on X, we can define a
relation ≡ on X given by:

x≡y if and only if gx = y for some g ∈ G

this clearly is an equivalence relation, due to the properties (P1)-(P3).


This relation have a huge importance on the groups, the equivalence classes of ≡
are known as the orbits of G, moreover we have a partition of X into orbits.

For example let’s consider the action of D4 into a square, we know that D4
have 8 elements, denoted as {e, a, b, c, r, s, t, u}. If we take a point x in the square,
Figure 4 shows the action of all elements into x.

r
s
e·x=x r·x

a·x b·x
t a
c

a·x b u·x

u u·x t·x

Figure 4: Action of D4

An action G is said to be transitive if there is only one orbit, for example,


if we consider the action of Cn into the vertices of the digraph of Figure 2. If
this doesn’t happen we say the action is intransitive. Note that for intransitive

10
actions, we have an action of G on each separate orbit that is, itself, transitive, as
for example thinking the action of D4 in x instead of the whole square. Moreover
if we want to describe all ways into a group can act in a set X it is sufficient to
describe transitive actions.

Now, if we want to describe the transitive actions, consider the following. Let
H be a subgroup, a left coset of H in G is the set gH = {gh : h ∈ H} for a fixed
g ∈ G. A classical, and useful, result is the following:
Theorem 4 (Lagrange’s Theorem). The H be a subgroup of G, if |H| = n and
|G| = m then n|m.
It is also important to remark the following:
Proposition 5 (Exercise 14.8.7). The number of left cosets and right cosets is the
same.
Proof. If ah ∈ aH, then (ah)−1 = h−1 a−1 ∈ Ha−1 . Now gH = g ′ H ⇐⇒ g −1 g ′ ∈
H. Since H is a subgroup, (g −1 g ′ )−1 = g ′−1 g ∈ H.
This implies that g ′−1 (g −1 )−1 ∈ H, which happens if and only if Hg −1 = Hg ′−1 .
Thus if ϕ(gH) = Hg −1 , we see that for any coset g ′ H = gH, that:
ϕ(g ′ H) = Hg ′−1 = Hg −1 = ϕ(gH), that is, ϕ is is independent of the repre-
sentative element. So ϕ is a well-defined function on left cosets. And clearly is a
bijection, so the result follows.
Now, we want to fact when two cosets are equivalent, for this, note that, if
g ′ ∈ gH, then g ′ = gh0 for some h0 ∈ H; then any element g ′ h ∈ g ′ H lies in gh
because g ′ h = g (h0 h) and h0 h ∈ H. Similarly, every element of g ′ H is in gH.
Then, if gH and g ′ H are not disjoint, g ′ H = gH.

Lagrange’s theorem tells us that the order of a subgroup divide the order of
the group, and that follows from noticing the number of cosets of H has the same
number of elements of H itself, i.e. the map h → g · h is a bijection from H to
gH. And moreover the number of cosets of H in G is |G|/|H| and this number we
write it as [G : H] and its called the index of H in G.

We can define the coset space (G : H) and the coset action of G on (G : H) is


given by:
g(kH) = (gk)H,
we can interpret this as, the permutation that correspond to g, maps kG to gkH
for all k ∈ G, and clearly is an action.
Theorem 6. Any transitive action G is equivalent to a coset action.

11
Proof. Let’s suppose that G acts transitively on X, and let x ∈ X, we can define
the set H = {g ∈ G : gx = x}, this is the subset of G that fixes x, and its known
as the stabilizer of x and is written as Gx . Note that Gx is a subgroup of G:
• The identity fixes x for any x ∈ X.
• If g ∈ Gx by definition gx = x hence g −1 gx = g −1 x and then x = g −1 x and
therefore g −1 ∈ Gx .
• If h, g ∈ Gx then (gh)x = g(hx) = gx = x and therefore gh ∈ Gx .
Now, there is a bijection between X and (G : H), defined as, let y ∈ X we can
define the subset S(y) = {g ∈ G : gx = y}, and as the action is transitive, we
know that this set is non-empty, and moreover, {S(y) : y ∈ Y } is a partition of X
and hence we can identify each partition with the partition in cosets given by H.
So this bijection define an equivalence in the actions of G over X, i.e. if gy = z
then g(S(y)) = g(S(z)), and this, by definition is a coset action.
Theorem 7. Two coset actions on (G : H) and (G : K) are equivalent if and only
if the subgroups H and K are conjugated.
Proof. Let’s suppose that H and K are conjugated, this is, there exists an ĝ ∈ G
such that K = gˆ−1 H ĝ, clearly, this implies the equivalence between the cosets
Kg → H ĝg.

For the the other direction let’s suppose that the two coset actions are equiva-
lent. Let’s suppose that the corresponding coset for K is H ĝ, then the stabilizers
of both are equal, the stabilizer of K is K itself, but the one for H ĝ is:
{g ∈ G : H ĝg = H ĝ} = g ∈ G : ĝgĝ −1 ∈ H = ĝ −1 H ĝ,


hence we get that K = ĝ −1 H ĝ.


The group actions make a distinction between labelled and unlabelled struc-
tures, if we consider an action of Sn on the class of labelled structures, unlabelled
structures corresponds to the orbits, and the stabilizers of an structure is its au-
tomorphism group, i.e. the set of all permutations fixing it.
Theorem 8. 1. The number of different labellings of a structure C is equal to
n!/| Aut(C)|
2. If there are M labelled structures and m unlabelled structures C1 , . . . , Cm ,
then m
X 1 M
=
i=1
|Aut (Ci )| n!

12
Schreier-Sims algorithm
A common problem is need to find the size of a group G, in this section we will
give an algorithm that can be used for this.

A consequence of Lagrange’s theorem, is that, if a group G acts on X, the size


of the orbit of X is equal to the number of cosets of the stabilizer Gx in G, and
this is more likely to be easy to calculate, as Gx is smaller than G. First, we need
to compute orbits, let S be a generating set for the group G acting on X. Consider
the following:

Algorithm to find orbit of x

Start with Y = ∅. Add the point x to Y .


While any point was added to Y in the previous step, apply all elements
of S to the recently added points, if you get a new point that is not in Y ,
add it.
Y is the orbit of x.

Let Y = {x = x1 , x2 , . . . , xs } be the orbit of x, and let ki maps x to xi , for


i = 1, . . . , s (with k1 = e ). If H = Gx , then k1 H, . . . , ks H are all the cosets of H
in G, in other words, k1 , . . . , ks are coset representatives for H in G.

Theorem 9 (Schreier’s Lemma). Let {g1 , . . . , gm } generate G; let k1 , . . . , ks be


a coset representatives for a subgroup H of G. We denote ḡ = ki if Hg = ki H.
Assume that k1 = e. Then H is generated by the set:
−1
n o
SH = ki gj (ki gj ) : i = 1, . . . , s; j = 1, . . . , m

As an example, let’s set G = S3 = {e, (1, 2), (1, 3), (2, 3), (123), (132)} and H
is a subgroup C3 = {e, (123), (132)}. We know that S3 is generated by s1 = (12)
and s2 = (123). We want to find a set of generators for C3 .

Note that C3 only has two cosets, and the coset representatives are t1 = e and
t2 = (1, 2). So:
t1 s1 = (12), so t1 s1 = (12),
t1 s2 = (123), so t1 s2 = e,
t2 s1 = e, so t2 s1 = e,
t2 s2 = (23), so t2 s2 = (12).

13
And:
−1
t1 s1 t1 s1 =e
−1
t1 s2 t1 s2 = (123)
−1
t2 s1 t2 s1 =e
−1
t2 s2 t2 s2 = (123)
As having e in the generator set is irrelevant, C3 is generated by (123).

This theorem is a way to find generators to a subgroup, in particular we can use


it to find the generators of the stabilisers. And this generators we will call them
as Schreier generators. Now, we can apply them recurrently to get the following:

Group Order: Schreier-Sins algorithm

Let S be a set of generators for G.


If S = ∅ or S = {e} then |G| = 1.
Otherwise, let x be a point not fixed by the elements of S, calculate the
orbit of Y of x and the Schreier generators of Gx . Apply the algorithm
recursively to |Gx |
Then |Gx | · |Y | = |G|.

This is useful calculating group size of big groups like the Rubik’s cube group,
but can be very hard to do it on hand, anyway let’s do a small example to see how
it works.

Example 10. Let’s consider a group G generated by x = (123), y = (124) and


H ≤ S4 .
We start with a single point 1 and the generating set S1 = {x, y}, note that:

e · 1 = 1,
x · 1 = 2,
x−1 · 1 = 3,
xy · 1 = 4.

Hence, the orbit of 1, is {1, 2, 3, 4} with permutations T1 = {e, x, x− 1, xy}. Now we


want to find the Schreier generators to the stabilizer G1 , calculating the products
on the way Theorem 9. Note that we get:

xx−1 = e,
x−1 y = (2, 4, 3).

14
as we get a non-identity element, we stop and adjoint a new base point 2 and add
a z = (2, 4, 3) as a generator, and using S2 = {z} for basepoint 2, note that:

e · 2 = 2,
z · 2 = 4,
z −1 · 2 = 3.

Hence, the orbit of 2 under S2 is {2, 3, 4} and now we want to find the Schreier
generators to the stabilizer G2 , it is easy to verify that all products given by the
Schreier lemma gives the identity and hence we stop the algorithm. And so we
know that the order of the group G is the multiplication of the basic orbit lengths,
i.e. 4 × 3 = 12.

15
Primitivity and multiple transitivity
We have reduced the study of group actions to only the ones that are transitive,
but it can be reduced even more. In general, a relation on X ban be thought as a
set of ordered pairs, i.e. is a subset of X × X.
Proposition 11. The relation R is perserved by G if and only if it is the union
of orbits of G in X 2 .
Proof. As the relation is G-invariant, this by definition is that if (x, y) ∈ R iff
(gx, gy) ∈ R, for any g ∈ G, and this implies that the whole orbit of (x, y) is in R,
hence R is the union of orbits. The other direction is similar.
Consider the following:
• A relation is preserved by G, or G-invariant, if x ∼ y implies that gx ∼ gy
and conversely.
• G acts on the set X × X as:
g(x, y) = (gx, gy).

• A G-congruence on X is an equivalence relation ∼ on X which is preserved


by G. We don’t require that G fixes the classes of equivalence of ∼.
• There always exists two trivial G-congruences:
1. The relation of equality, i.e. x ∼ y iff x = y.
2. The all relation, i.e. x ∼ y for all x, y ∈ X.
• If there are more than this two trivial G-congruences, the group is said to
be imprimitive, and otherwise, primitive.
As an example, consider S4 acting on the set {1, 2, 3, 4} and the permutation:
 
1 2 3 4
σ=
2 3 4 1
The group generated by σ is not primitive. To see this, consider the partition
(X1 , X2 ) of X given as X1 = {1, 2} and X2 = {2, 3}; note that σ(X1 ) = (X2 ) and
σ(X2 ) = X1 , i.e. the action of σ on the partition gives an non-trivial equivalence
relation over X, and hence, is imprimitive.

Let G be a transitive permutation group, if ∼ is a non-trivial G-congruence,


consider X1 , . . . , Xm be the congruence classes, and Y = {X1 , . . . , Xm } the set of
classes; we can define two new permutation groups:

16
• G acts on the set Y and let G0 be the permutation group on Y induced by
G.

• Let H be the subgroup of G that fixes X1 as a set, and H0 be the permutation


group induced by X1 by H.

Theorem 12. G is isomorphic to a subgroup of the wreath product H0 ≀ G0 .

This tells us that G can be constructed out of smaller groups which are both
transitive, and if they are imprimitive, we can recursively continue until we end
up with a collection of primitive subgroups, known as primitive components of G,
which are not necessarily unique, as G may have different congruences.

Now, let take t ≤ |G|, a permutation group G is said to be t-transitive if,


for any t-tuples (x1 , . . . , xt ) and (y1 , . . . , yt ) of distinct points of X, there exists
a permutation g ∈ G such that gxi = yi for any i. Clearly t = 1 is the same as
transitivity.

Proposition 13. If G acts t-transitively on X for t ≥ 2, then:

1. G is (t − 1)-transitive,

2. G is primitive.

Proof. 1. Let’s consider two (t − 1)-tuples, (x1 , . . . , xt−1 ) and (y1 , . . . , yt−1 ), we
can extend them to t-tuples by adding elements that are not in the tuples xt
and yt respectively. As G is t-transitively, there is an g such that gxi = yi
for i = 1, . . . , n. And then we can use the same g for the (t − 1)-tuples.

2. Due to part 1 we can consider G to be 2-transitive. and Proposition 11 tells


us that any G-congruence is the union of orbits acting on X × X, and this
necessarily contains the diagonal ∆ = {(x, x) : x ∈ X}, as any G-congruence
is in particular an equivalence relation and, therefore, reflexive.
But, as G is 2-transitive, it has only two orbits on X 2 , namely ∆ and X 2 \∆;
so, there are only two possible congruences, and hence, G is primitive.

It is known that Sn acting on {1, . . . , n} is n-transitive for any n and the alter-
nating group An acting on the same set is (n − 2)-transitive for n > 2. And this
implies that Sn is primitive for any n and An is primitive for any n > 2.

Another very interesting fact is that no other finite permutation group can be
more than 5-transitive, as a consequence of the classification of the finite simple
groups.

17
Example
Let’s consider the Petersen Graph:

(1, 2)

(4, 5) (3, 5) (3, 4)

(1, 3) (2, 5)

(1, 4) (2, 4)

(2, 3) (1, 5)

Figure 5: Petersen graph

It can be proved that the number of automorphisms of this graph is 120. Now,
we have labeled each vertex with a 2-tuple of the set {1, 2, 3, 4, 5}, so, there are
5
2
= 10 possible 2-subsets that are used. Note also that two vertices are adjacent
if and only if their labels are disjoint, hence any permutation of {1, 2, 3, 4, 5} in
the inducted action over these pairs is an automorphism. It follows that, as we
find a group of automorphism isomorphic to S5 on X × X with order 120, so the
full automorphism group is S5 .

Is easy to note that the automorphism group is vertex transitive, but is not 2-
transitive, since no automorphism map two adjacent vertices to two non-adjacent.
However, the orbits of S5 on X 2 are three:

• the diagonal {(x, x) : x ∈ X};

• the set {(x, y) : x ∼ y};

• the set {(x, y) : x ̸= y, x ≁ y}.

The automorphism group is transitive on (ordered) edges and on (ordered) non-


edges. So, now we can show that S5 is primitive on X, as a congruence R must
be the union of some of three orbits, but off course including the diagonal.

18
Suppose that the relation contains the second orbit (that can be interpretated
as the ordered edges) and not the third one (that can be interpretated as the
ordered non-edges). Since we can find vertices x, y, z such that x ∼ y ∼ z, by
transitivity of the relation x ∼ z but then this imply that the relation include also
the third orbit. A similar argument works if we suppose that the relation include
the third but not the second orbit.
So, either, the relation is the identity, or the whole X 2 , and therefore, the group
is primitive.

19
Cayley Digraphs and Frutch’s Theorem
Let S be a subset of a group G, not containing the identity, we can define the
Cayley digraph of a group with respect to S.

Definition 14. Let G be a finitely generated group and S a subset of G, we can


define the Cayley digraph ΓG,S , which is a directed graph that can be constructed
following the next steps:

1. Each g ∈ G is a vertex of vg ∈ V (ΓG,S )

2. Each s ∈ S forms a directed edge with initial vertex vg and terminal ver-
tex vg·s ; Giving each edge a correspondence with right multiplications of the
elements of S.

Note for example, that Figure 2 is exactly the Cayley digraph of Cn with re-
spect to the generating set S = {1}.

Cayley digraphs can change a lot depending on the set S, for example, let’s
consider the group S3 with S1 = {(12), (23)} and S2 = {(12), (123)}. The first one
is illustrated in Figure 6b, and the other in Figure 6a. In those graphs, to avoid
the use of cycles, we will use the double arrows to simplify the graph.

(12) (123)
(12)

e
e (13)

(132) (123)
(23) (132) (23) (13)
(a) S1 = {(12), (23)}. (b) S2 = {(12), (123)}.

Figure 6: Cayley graphs of S3 with different generators

Proposition 15. 1. The Cayley digraph ΓG,S of a group G with respect to S is


connected if and only if S generates G.

2. For each g ∈ G, the map ρg : x 7→ gx is an automorphism of ΓG,S .

20
Proof. 1. If S generates G then every element g ∈ G can be written as the
product of the elements of S, in the Cayley digraph this can be see as a path
from e to g, it follows that ΓG,S is connected, but not necessarily strongly
connected.
The converse is pretty similar, any path from e to g in the underlying graph
translates into a product of elements of S that equals g, it follows that S
generates G.
2. Let (x, sx) an edge, then (ρg x, ρg sx) = (gx, gsx) is also an edge, and the
result follows.

Proposition 16 (Exercise 14.7.12). If G is finite and S is a generating set of G,


then ΓG,S is strongly connected.
Proof. If G is finite then the generator have finite order, then the Cayley digraph
is composed of cycles, so is strongly connected.
If G is not finite, for example, in the Cayley digraph of the free group on n
generators with respect to those free generators, there is no path to the vertex
corresponding to the identity from the vertex corresponding to any one of those
generators.

We have an action of G on the vertices of the Cayley digraph, as a group of


automorphisms. Note that this action is transitive; for ρg−1 h maps g 7→ h, so we
denote the permutation group ρ(G), so we think on ρ as the action of G.
Proposition 17. Suppose that S generates G. Then any automorphism of ΓG,S
which preserves the labels on the edges belongs to ρ(G).
Proof. Let f be an automorphism which preserves the labels. Since all elements
of p(G) also preserve labels, we can compose f with the element ρf −1 (1) to obtain
an automorphism fixing 1 ; and this automorphism lies in ρ(G) if and only if f
does. So we may assume that f fixes 1 . Now, for each s ∈ S, there is a unique
edge with label s and initial vertex 1 (namely (1, s)), and a unique edge with label
s and terminal vertex 1 (namely (s−1 , 1) ). So f must fix all elements s or s−1
for s ∈ S. In this way we can work out through the digraph, and find that f
fixes every element which is a product of elements of S and their inverses. But,
by assumption, these elements comprise all of G; so f = 1 ∈ ρ(G).
Now, we get the following result:
Theorem 18 (Frutch’s Theorem). Every finite group is the automorphism group
of a finite graph.

21
References
[1] Peter J. Cameron, Combinatorics: Topics, Techniques, Algorithms, Cambridge University
Press, 1995.
[2] Jhon Meier, Groups, Graphs and Trees. An Introduction to Geometry of Infinite Groups,
London Mathematical Society, 2008.

22

You might also like